Oecologia (2006) 147: 24–30
DOI 10.1007/s00442-005-0172-7
POPULATION ECOLOGY
Leif Egil Loe Æ Christophe Bonenfant Æ Rolf Langvatn
Atle Mysterud Æ Vebjørn Veiberg Æ Nils Chr. Stenseth
Increased effect of harsh climate in red deer with a poor set of teeth
Received: 21 January 2005 / Accepted: 23 May 2005 / Published online: 10 December 2005
Springer-Verlag 2005
Abstract Teeth are vital for mammal performance and
especially in ungulates relying on mechanical decomposition of plant material for effective microbial digestion
and energy uptake. The main focus of the role of teeth in
ungulate life histories has been on tooth wear, while no
one has addressed to what extent deviation from the
natural set of teeth (maldentition) causes variation in
individual fitness components. Based on mandibles from
41,066 individual red deer (Cervus elaphus L.) collected
from 1969 to 2001, we tested whether maldentition had
an effect on individual body condition and whether this
effect depended on environmental harshness. Females
with maldentition (0.6% of the population) were in a
poorer condition than individuals without tooth anomalies and the effect increased during unfavorable climatic
conditions. The effect of maldentition in males was less
clear. This study indicates that a well-functioning set of
teeth is essential for mammal performance, and that
selection pressure against (dental) anomalies is more
pronounced when climate is unfavorable.
Keywords Dentition Æ NAO Æ Ungulates Æ Body
weight Æ Mechanisms
Electronic Supplementary Material Supplementary material is
available for this article at http://dx.doi.org/10.1007/s00442-0050172-7
L. E. Loe Æ C. Bonenfant Æ A. Mysterud (&) Æ V. Veiberg
N. Chr. Stenseth
Centre for Ecological and Evolutionary Synthesis (CEES),
Department of Biology, University of Oslo, P.O. Box 1066,
Blindern, 0316 Oslo, Norway
E-mail: [email protected]
Tel.: +47-22-854045
Fax: +47-22-854726
R. Langvatn Æ V. Veiberg
The University Centre in Svalbard (UNIS),
9170 Longyearbyen, Spitsbergen, Norway
R. Langvatn
Norwegian Institute for Nature Research,
Tungasletta 2, 7005 Trondheim, Norway
Introduction
Mammals depend on a well-functioning set of teeth to
acquire and process food. This is particularly critical for
ungulates where effective energy absorption in the
digestive tract depends on the molariform teeth’s ability
to reduce plant particle size to ensure more efficient
microbial digestion (Pérez-Barberı́a and Gordon 1998).
Size and form of the front teeth (in addition to muzzle
width) may also affect a herbivore’s ability to select the
preferred forage plants (Gordon and Illius 1988). Recent
studies have investigated the importance of tooth wear
for the life history of ungulates (Skogland 1988; Ericsson
and Wallin 2001; Loe et al. 2003). Maldentition (i.e.,
deviation from the natural set of teeth) is likely to reduce
teeth function to a higher extent than wear. Decreased
longevity was found in elk Cervus elaphus with severe
tooth damages due to anomalous exposure to fluoride
and silica (Garrott et al. 2002). Only two large-scale
studies (Zima 1988; Azorit et al. 2002) have reported
frequency of maldentition occurrence in natural populations of ungulates (but see, e.g., Kierdorf and Kierdorf
2002 for single individuals) and no study has so far
investigated the consequences of tooth anomalies for
ungulate life history under more normal conditions
(than those present in the study system of Garrott et al.
2002).
In general, the environment affects life history traits
(such as condition, fecundity, and survival) in ungulates
(Coulson et al. 2000), but inferior and injured individuals may suffer even more from environmental stress
than ‘normal’ individuals. In the specific case of maldentition, individuals with inferior teeth may suffer more
at high population density and/or under adverse climatic
conditions (especially if there is a deep snow layer), if
maldentition were related to their poorer ability to
masticate sub-optimal diets composed of rougher items.
Anomalies may accumulate throughout life and become
progressively more important for individual performance as they age.
25
Here, we report on the age- and sex-specific frequency
of maldentition in a sample of 41,066 Norwegian red
deer (Cervus elaphus L.). We check if the frequency of
tooth anomalies increases by age. Thereafter, we test for
the first time the hypotheses that body condition is lower
in red deer with maldentition, and that this is more
pronounced at a high population density and during
unfavorable climatic conditions.
Materials and methods
Study area
The study area consists of 65 municipalities covering
large parts of the Norwegian south–west coast, which is
the main area for red deer in Norway (Langvatn et al.
1996). Red deer in the study area can be divided into five
populations corresponding to Norwegian counties:
Rogaland and Hordaland (P1), Sogn og Fjordane (P2),
Møre og Romsdal and Sør-Trøndelag (P3), NordTrøndelag (P4), and the island Hitra (P5; Mysterud et al.
2000). Red deer eat graminoids and herbs in summer
and prefer blueberry and cowberry (Vaccinium L. sp.) in
the winter. However, when the snow layer is deep they
no longer have access to these dwarf shrubs and change
to browse and bark stripping in the tree layer (aspen
(Populus tremula L.), rowan (Sorbus aucuparia L.), willow (Salix L. sp.), juniper (Juniperus communis L.), and
Scots pine (Pinus sylvestris L.) (Ahlén 1965; Ahlén
1975).
post-mortem changes were distinguished from antemortem ones by looking at the alveoli and by the condition of the tooth fracture planes.
Environmental covariates
In an average year, snow covers the ground from
December to March; however, snow cover is dependent
on altitude and phase of the North Atlantic Oscillation
(see Hurrell and Van Loon 1997 for definition of the
NAO). On the west coast of Norway, high NAO values
in winter correlate with high temperatures and much
precipitation (falling mainly as rain at low elevation, and
more often as snow at high elevation). In lowland red
deer wintering areas, a significant snow cover is only
present in years with low NAO value (Mysterud et al.
2000). We used the mean NAO value from December to
March as a covariate in our analyses. Recently, the use
of large-scale weather packages (such as the NAO) has
been acknowledged to predict ecological processes better
than local climate in some cases (Hallett et al. 2004).
In order to obtain an index for population density in
a given year and municipality, we divided the total
number of harvested red deer on the ‘qualifying area’,
which is the area defined as deer habitat by the management authorities (see Mysterud et al. 2001c for details).
Statistical analyses
Frequency of tooth anomalies
Red deer data
A total of 41,066 mandibles from adult red deer
(>1 year old) were sampled during the annual autumn
harvest 1969–2001 (between 10th September and 15th
November), together with data on body weight and records of date of culling and locality (municipality)
(Langvatn et al. 1996). Recorded body weight is dressed
weight (accounting for 58% of live weight; Langvatn
1977). Jaw length was measured according to Langvatn
1977). Age was estimated from annual tooth cementum
layers (Mitchell 1967; Reimers and Nordby 1968).
Dental abnormalities were classified according to the
involved teeth and the malformation type (Table 1). The
various types of maldentition are a combination of
innate anomalies and mechanical damages inflicted
throughout life. Due to the relatively low number of
individuals with tooth anomalies, we did not attempt to
analyze subsets of specific types of teeth nor specific type
of anomalies (listed in Table 1). We found that the origin and the age of a tooth anomaly (i.e., if it was innate
or inflicted) could not be established for the majority of
individuals. Hence, we do not separate ‘innate’ or
‘inflicted’ tooth anomalies in the following analyses.
Teeth can sometimes be lost from cleaned, dry jaws, or
fracture during improper handling of mandibles. These
Tooth anomaly occurrence was modeled as a binary
response (present or not) using the Generalized Linear
Model with a logit link and binomial error distribution
(McCullagh and Nelder 1989; Agresti 1990). However,
given the relatively low occurrence of maldentition in
red deer (>1% in average), standard logistic regression
might give biased results (McCullagh and Nelder 1989;
Venables and Ripley 1999) by underestimating the
probabilities of the studied event to happen (King and
Zeng 2001). A way to overcome such biases is to modify
the maximum likelihood estimator (Firth 1993; Heinze
and Schemper 2002), which bears the consequence that
we could not compute the Akaike Information Criterion
(AIC; Akaike 1974) required for the theoretical information approach as we did in the body mass analyses
included in this paper. As recommended for the modeling of rare events such as tooth anomalies, profile
likelihoods were used to test the effect of year, population, sex, and age on tooth anomaly occurrence (Venables and Ripley 1999; Heinze and Schemper 2002).
Finally, we assessed the model fit (goodness-of-fit tests,
GOF) according to Agresti (1990) by (1) transforming
continuous variables into an 8-class categorical variable
and (2) comparing the Pearson residuals with a Chisquare with the residual number of degrees of freedom.
26
Table 1 Occurrence of the different types of maldentition in molars, premolars, and incisiforms of female and male Norwegian red deer
Molars
Females
No maldentition
Missinga
Brokenb
Malocclusion (incl torsion)
Supernumerary teeth
Not specified/Other
Males
No maldentition
Missinga
Brokenb
Malocclusion (incl torsion)
Supernumerary teeth
Not specified/Other
a
Premolars
Incisors
Total
n
Percentage
n
Percentage
n
Percentage
n
Percentage
16,773
5
0
4
1
12
99.87
0.03
0.00
0.02
0.01
0.08
16,735
38
1
6
1
13
99.65
0.23
0.01
0.04
0.01
0.08
16,760
5
10
9
1
2
99.84
0.03
0.06
0.05
0.01
0.02
16,735
48
11
19
3
27
99.36
0.28
0.07
0.11
0.02
0.16
24,134
3
2
1
99.93
0.01
0.01
0.00
0.00
0.04
24,104
23
2
4
5
17
99.79
0.10
0.01
0.02
0.02
0.07
24,104
6
5
24
1
15
99.79
0.02
0.02
0.10
0.00
0.06
24,104
32
9
29
6
43
99.51
0.13
0.04
0.12
0.02
0.18
11
No visible remains
Visible remains of the tooth
b
We assume that a good fit of simple models (including
only one predictor variable) indicates that extended
models also fit well (as done in Langvatn et al. 2004). All
GOF P-values were >0.4, suggesting that the models
captured the data adequately.
studies based on observational data (Johnson and Omland 2004), although it inflates the standard errors of
parameter estimates. An example of this model averaging procedure in ecology is given by Conner et al. (2001).
From model AICs, we derived the relative AIC weights
as
Body weight analyses
e0:5DAICi
^ i ¼ Pk
w
;
0:5DAICj
j¼1 e
The effects of maldentition on log-transformed (to stabilize variance) body weight were investigated with the
Generalized Linear Model (McCullagh and Nelder
1989) using an identity link function and a normal error
distribution. All factors shown earlier to affect red deer
body weight were included in our initial model (see
Yoccoz et al. 2002; Loe et al. 2003 for a similar approach). Theses variables were age (as 5 age classes for
females and a 6-order polynomial function for males;
Mysterud et al. 2001c), density (Mysterud et al. 2001c),
the NAO (Mysterud et al. 2001b), latitude (Mysterud
et al. 2001a), diversity of aspect (with the ShannonWiener information criteria; data: proportion of area
facing north, northeast, east, southeast, south, southwest, west, and northwest in each municipality), proportion of high altitude habitat and the distance to the
coast (Mysterud et al. 2001a), and population (Mysterud
et al. 2001a). Second-order interactions between population, age, and the NAO were also entered in the model.
We followed the same procedure as Yoccoz et al. (2002)
and standardized covariables so that coefficients could
be easily interpreted as well as allowing the comparison
of the relative strength of each covariable.
The model selection was based on the Akaike Information Criterion (AIC), which selects the most parsimonious model. We followed the information theoretic
and model averaging approaches (Burnham and
Anderson 1998) which include more than one single best
model. This procedure has recently been promoted as a
useful tool in evolutionary biology and ecology for
which gives the probability for the model i to be the best
model. This model averaging procedure thus accounts
for the model selection uncertainty that is an integrated
part of the statistical inference (Buckland et al. 1997).
We considered a set of best models including all models
within a range of four AIC points from the best model to
build up the average model. The average model estimates were estimated as the weighted mean (^
wi ) of the
set of best model coefficients. Coefficient variances were
calculated as
varbj ¼
k
X
^i
w
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
varðbj;i Þ þ ðbj þ bÞ2 ;
i¼1
where bj is the coefficient of interest and bj,i the estimate
of bj knowing model i (Burnham and Anderson 1998).
According to Burnham and Anderson (1998), these
average model estimates of the coefficients are lowered
to 0 and have increased variance when they do not appear in the ith model. Hence, some variables may be
selected in the average model without being significantly
different from 0.
We assessed independent probability for the ith
model to be the best model using a bootstrap resampling
method (Efron and Tibshirani 1993) denoted as pi
(Buckland et al. 1997; Burnham and Anderson 1998) by
analyzing resamples and selecting the model as done in
the original data. We ran 10,000 bootstrap batches
27
(treating each observation as independent) to test the
reliability of the standard AIC model selection.
As we used log body weight as the response variable
in all analyses, one may argue that the causal relationship between maldentition and phenotypic quality cannot be established in transversal data when using body
weight as response. Individuals may become small as a
result of tooth anomalies, or they may get tooth
anomalies as a consequence of their lower phenotypic
quality. We therefore included a measure of skeletal
body size (jaw length) as a predictor so that the effect of
maldentition on body weight after skeletal body size is
accounted for can be interpreted as an effect on body
condition. Body condition is much less likely to influence maldentition than the other way around, avoiding a
discussion of causality in our results. However, we also
performed the analyses without including jaw length and
this gave qualitatively the same results.
All analyses were performed in S-Plus (Insightful,
Seattle, WA, USA) and R (R Development Core Team
2004).
Results
The overall frequency of any type of maldentition in
Norwegian red deer was 0.6% for females and 0.5% for
males (Table 1). The frequency of maldentition increased by age both in females and males (from less than
0.5% in young age classes to more than 6% in the oldest
females; Fig. 1), indicating that most of the damages in
the old individuals are not innate, but have accumulated
throughout life. There was no main effect of sex (estimate=0.3411, SE=0.2670) nor an interaction between
sex and age (estimate=0.0318, SE=0.0293). Probability
Fig. 1 The proportion of individuals with tooth error by increasing
age in red deer. Since there is no sex effect, the predicted model is
the same for males and females. The area of data points is
proportional to sample size. Note that we have no males in the
oldest age group (older than 20 years). For the young age classes,
standard errors are so small that they are covered by the data
points. The y-axis in females was cut for clarity (i.e., the upper part
of the standard error bar of >20-year-old females was cut, but is
similar to the lower part)
for tooth anomalies did vary in space (populations P3
[estimate=0.6974, SE=0.1935] and P5 [estimate=
1.033, SE=0.317] have lower probability of maldentition than the reference population P1). There was no
significant temporal trend in tooth anomalies occurrence
over the course of the study (estimate=0.0091,
SE=0.0083).
As reported earlier, the NAO positively affected body
weight of red deer when the index is above 2 (reversed
below this level; see Mysterud et al. 2001b), and the
effect is strongest for positive NAO values (slightly nonlinear). Further, novel to this study, the effect of the
NAO was stronger in female deer with maldentition
(Fig. 2; Tables 2 and 3; see electronic appendix S1 for
the full set of models). Females with maldentition were
on average 4 kg lighter than ‘normal’ females during
harsh climate (NAO around 1) while the weight did
not differ when climate was favorable (NAO around 4;
Fig. 2; all other factors, also skeletal size, accounted
for). The effect of maldentition was not stronger at high
densities (the interaction term density · maldentition
was not selected in the best models; Table 2). In males,
there was no clear effect of maldentition on body condition, although the pattern was qualitatively consistent
with the pattern in females (i.e., the parameter estimates
for the mean effect were similar, but the SE was larger;
see electronic appendix S2).
Discussion
Our study reports for the first time consequences of
maldentition on body condition, which is a key life
history trait in ungulates. Maldentition was rare (0.6%),
Fig. 2 Influence of climate (as given by winter NAO) on body
weight of female red deer with and without maldentition, adjusted
for all other factors (including jaw length, which is a measure of
skeletal body size). Climate is most adverse around NAO values
close to 2 (see Mysterud et al. 2001b for the relationship between
NAO and local climate parameters). The (back-transformed
logarithmic) predicted values obtained with the parameters from
the average model (see Materials and methods) were used to draw
the figure
AIC refers to Akaikes Information Criterion (Akaike 1974), D AIC is the difference in AIC value between the given and the best model, wi is the theoretical probability for the given
model to be the right one (wi of all a priori models sum to 1). See Materials and methods for details. Terms related to maldentition and the interaction between maledentition and NAO
are bolded
4.2
13954.18
to coast + population + maldentition
0.048
3.22
13955.16
to coast + population + maldentition
0.079
1.47
13956.91
to coast + population + maldentition
0.189
0.94
13957.44
0.246
0
13958.38
to coast + population + maldentition
age·density
to coast + population + maldentition
ln(body weight)ln(jaw length) + age + density + diversity of aspects + date of culling + distance
+ NAO + NAO2 + NAO3 + NAO·maldentition + NAO2·maldentition + NAO3·maldentition +
ln(body weight)ln(jaw length) + age + density + diversity of aspects + date of culling + distance
+ NAO + NAO2 + NAO3 + NAO·maldentition + NAO2·maldentition + age·density
ln(body weight)ln(jaw length) + age + density + diversity of aspects + date of culling + distance
+ NAO2 + NAO3 + NAO·maldentition + NAO2·maldentition + age·density
ln(body weight)ln(jaw length) + age + density + diversity of aspects + date of culling + distance
+ NAO + NAO2 + NAO3 + NAO·maldentition + NAO2·maldentition + NAO3·maldentition
ln(body weight)ln(jaw length) + age + density + diversity of aspects + date of culling + distance
+ NAO + NAO2 + NAO3 + NAO·maldentition + NAO2·maldentition
0.393
D AIC
AIC
Model females
Table 2 The five best models for ln(body weight) in female Norwegian red deer (see electronic appendix S1 for all models)
wi
28
but had a profound negative effect on body weight in
females when the climate was unfavorable.
Too few studies have been reporting population
frequencies of maldentition to know if it is generally
rare among long-lived mammals. The single earlier
study reporting the frequency of maldentition in red
deer provides a remarkably similar estimate (0.8%
males and 0.6% females had abnormal teeth, n=1,091
individuals) for a population on the southern border of
the distribution range (in Spain; Azorit et al. 2002). In
roe deer, the frequency of numerical tooth anomalies
was much higher (4.85%; n=16,510 upper jaws and
16,177 lower jaws) but did not vary among 14 districts
within Czechoslovakia (Zima 1988). Deviations from a
normal set of teeth are known to be relatively common
in this species (Zima 1988), suggesting that the frequency of maldentition within ungulates is highly species specific.
We find that maldentition occurs more than ten times
more often in old than in young females, indicating an
accumulation throughout life (see Kratochvil 1984 for a
similar pattern in roe deer). The predicted value is much
lower than the observed value for the oldest age group
(Fig. 1), indicating a rapidly accelerating deterioration
of the teeth as the individuals reach their terminal age.
This is in line with the observed pattern of reproductive
(Langvatn et al. 2004) and phenotypic (Mysterud et al.
2001c) senescence in this population. It is likely a strong
selection for a set of teeth that lasts for a normal lifetime, but not much longer. However, tooth anomalies
are not the main factor causing senescence in this population since it occurs in <10% of females that are close
to terminal age. Since the frequency of tooth anomalies
increased by age, the population frequency of maldentition will rely heavily on the age structure. Norwegian red deer are harvested extensively and age structure
is biased towards young individuals, especially in males
(Langvatn and Loison 1999).
The rarity of maldentition, especially in young individuals, may be due to strong selection against innate
tooth anomalies (some likely genetic in origin). We here
demonstrate that such a selection pressure is more severe
during unfavorable environmental conditions. In a
broader context, this implies that individuals either born
with or with a genetic susceptibility to acquire even
minor anomalies are punished by external factors to a
higher extent than ‘normal’ individuals under unfavorable conditions. The rarity of the phenomenon implies
that maldentition per se is of little importance in population dynamics, especially in harvested populations
with a high proportion of younger individuals.
The declining function of teeth by age has a central
role in the discussion of proximate factors for senescence
in ungulates (Klein and Olson 1960; Gaillard et al. 1993;
Loe et al. 2003). Garrott et al. (2002) linked severe
malformation of teeth to shorter lifespan in elk. In Loe
et al. (2003), we found no relationship between molar
height (a proxy for tooth wear) and body condition.
Here, we demonstrate that in the same populations,
29
Table 3 Parameter estimates
(including 95% CI) of the
average model (see Materials
and methods) for body weight
in female red deer
Age categories are 1 2 years old,
2 3 years; 3 4 years, 4 5–20
years (prime age), 5 more than
20 years old (senescent). The
effects of age categories 1–4 are
shown by contrasting them with
age category 5. The same
approach is taken on the effect
of population (P1 to P5). Parameter estimates with 95%CI
not overlapping with 0 are
significant on P<0.05 level.
Terms related to maldentition
and the interaction between
maledentition and NAO are
bolded
Intercept
ln(jaw length)
Age category (1–5)
Age category (2–5)
Age category (3–5)
Age category (4–5)
Density
Diversity of aspects
Date of culling
Distance to coast
Population (P1–P5)
Population (P2–P5)
Population (P3–P5)
Population (P4–P5)
Maldentition
NAO
NAO2
NAO3
Maldentition·NAO
Maldentition·NAO2
Maldentition·NAO3
Age category 1·density
Age category 2·density
Age category 3·density
Age category 4·density
maldentition, which decreases tooth function to a higher
extent than normal wear, has a strong effect on condition for the females that are affected.
Male body condition was not affected by tooth
anomalies to the same extent as females (electronic
appendix S2). For each sampled individual, tooth
anomalies became effective at an unknown time between
the eruption of the permanent teeth (1–2 years of age
dependent on the tooth; Loe et al. 2004) and age
at harvest. Due to sex-biased harvest (Langvatn and
Loison 1999), sampled males with tooth anomalies were
younger (mean=4.41, median=3 years old) than
females (mean=9.58, median=8 years old). If there had
been a cumulative effect of tooth anomalies, the lacking
condition effect in males could be because they (on
average) have been subjected to maldentition for a
shorter time period than females. However, we could not
detect such a cumulative effect in our analyses (the
interaction term age · maldentition was not selected in
the best models). Maldentition may also interact with or
speed up the rate of tooth wear, which may potentially
come into play for the older females, although we did
not find a body condition effect of normal wear in Loe
et al. (2003). In general, body weights of males
(CV=24%) are much more variable than those of females (CV=15%), which may mask a minor effect of
maldentition in males. To detect a significant effect of
tooth anomalies on body weight would therefore require
a larger data set for males than for females (they are now
of similar size; Table 1), implying that the male data set
may be marginal. Indeed, as the parameter estimates for
the mean effect were similar for males and females (but
the SE was larger for males), lower power for males may
likely be the case.
Parameter estimate
95% CI
8.185495393
2.01
0.0367
0.0165
0.013
0.0095
0.035
0.501
0.000393
0.0375
0.0195
0.0164
0.00634
0.0059
0.072
0.00278
0.000154
8.80E-05
0.00735
0.00599
0.00063
0.0194
0.0111
0.00705
0.00297
8.84, 7.53
1.92, 2.09
0.0736, 0.000231
0.00241, 0.0307
0.00470, 0.0213
0.00514, 0.0139
0.0522, 0.0178
0.254, 0.749
0.000495, 0.000292
0.0316, 0.0434
0.0163, 0.0228
0.0143, 0.0185
0.0138, 0.00113
0.00839, 0.00340
0.105, 0.0395
0.000554, 0.00501
0.000278, 0.000586
0.000252, 7.61E-05
0.0145, 0.0292
0.00170, 0.0103
0.00226, 0.00100
0.0205, 0.0592
0.0267, 0.00446
0.0163, 0.00220
0.00773, 0.00178
The effect of tooth anomalies on female body condition was dependent on climate, but not on population
density. Adverse climate and population density are two
forms of environmental stress that are known to affect
herbivore performance. During winter, red deer prefer
blueberry and cowberry in the field layer and only shift
to browse in the tree layer if the former is not available
(Ahlén 1965, 1975). Therefore, severe climate and high
density would intuitively lead to the same effect, namely
a shift towards rougher type of diets due to, respectively,
snow coverage and reduction of easily chewable forage
in the field layer. We identified an overall negative effect
of population density (i.e., a main effect) in agreement
with earlier work (e.g., Mysterud et al. 2001b), but no
interaction between maldentition and density was detected. A likely explanation for why female deer with
maldentition are more susceptible to unfavorable climate than to high density is that deer habitats per date
are not grazed to such an extent that would cause a
sufficiently large shift in diet types. Along the west coast
of Norway, winters with much snow at high altitudes
(summer range) coincide with less snow at low altitudes
(winter range) (Mysterud et al. 2000). The NAO is
known to at least partly operate through summer foraging conditions, as it affects snow accumulation and
hence plant phenology and foraging conditions during
early summer at high altitudes (Mysterud et al. 2001b).
If snow cover totally inhibits access to the field layer,
this would force deer to eat a lower quality diet and
make tooth anomalies fatal. As it is more difficult to see
how maldentitions can affect performance from the
summer grazing situation, our study therefore also
provides some indirect evidence that the NAO may also
operate through winter conditions.
30
Acknowledgements We gratefully acknowledge the financial support provided by the AURORA grant from the Research Council
of Norway (NFR) and EGIDE (French Ministry for Foreign
Affairs) to L.E.L and C.B. We thank David Firth, Nigel G.
Yoccoz, Torbjørn Ergon, and Jon Olav Vik for valuable statistical advice, and Jean-Michel Gaillard and the two anonymous
referees for valuable comments on an earlier draft. We declare
that no action associated with this work violates the current laws
of Norway.
References
Agresti A (1990) Categorical data analysis. Wiley, New York
Ahlén I (1965) Studies on the red deer, Cervus elaphus L., in
Scandinavia. Swed Wildl Res 3:177–376
Ahlén I (1975) Winter habitats of moose and deer in relation to
land use in Scandinavia. Swed Wildl Res 9:45–192
Akaike H (1974) A new look at statistical model identification.
IEEE T Automat Contr AU 19:716–722
Azorit C, Munõs-Cobo J, Analla M (2002) Abnormal teeth in the
Spanish red deer. Z Jagdwiss 48:252–260
Buckland ST, Burnham KP, Augustin NH (1997) Model selection:
an integral part of inference. Biometrics 53:603–618
Burnham KP, Anderson DR (1998) Model selection and inference:
a practical information-theoretic approach. Springer, Berlin
Heidelberg New York
Conner MM, White GC, Freddy DJ (2001) Elk movement in response to early-season hunting in northwest Colorado. J Wildl
Manage 65:926–940
Coulson T, Milner-Gulland EJ, Clutton-Brock T (2000) The relative roles of density and climatic variation on population
dynamics and fecundity rates in three contrasting ungulate
species. Proc R Soc Lond Ser B 267:1771–1779
Efron B, Tibshirani RJ (1993) An introduction to the bootstrap.
Chapman and Hall, London
Ericsson G, Wallin K (2001) Age-specific moose (Alces alces)
mortality in a predator-free environment: Evidence for senescence in females. Ecoscience 8:157–163
Firth D (1993) Bias reduction of maximum likelihood estimates.
Biometrika 80:27–38
Gaillard J-M, Delorme D, Boutin J-M, Van Laere G, Pradel R
(1993) Roe deer survival patterns: a comparative analysis of
contrasting populations. J Anim Ecol 62:778–791
Garrott RA, Eberhardt LL, Otton JK, White PJ, Chaffee MA
(2002) A geochemical trophic cascade in Yellowstone’s geothermal environments. Ecosystems 5:659–666
Gordon IJ, Illius AW (1988) Incisor arcade structure and diet
selection in ruminants. Funct Ecol 2:15–22
Hallett TB, Coulson T, Pilkington JG, Clutton-Brock TH, Pemberton JM, Grenfell BT (2004) Why large-scale climate indices
seem to predict ecological processes better than local weather.
Nature 430:71–75
Heinze G, Schemper M (2002) A solution to the problem of separation in logistic regression. Stat Med 21:2409–2419
Hurrell JW, Van Loon H (1997) Decadal variations in climate
associated with the North Atlantic Oscillation. Clim Change
36:310–326
Johnson JB, Omland KS (2004) Model selection in ecology and
evolution. Trends Ecol Evol 19:101–108
Kierdorf U, Kierdorf H (2002) Supernumerary incisiform tooth in
a red deer. Z Jagdwiss 48:276–280
King G, Zeng L (2001) Explaining rare events in international
relations. Int Organ 55:693–715
Klein DR, Olson ST (1960) Natural mortality pattern of deer in
southeast Alaska. J Wildl Manage 24:80–88
Kratochvil Z (1984) Verandrungen am Gebiss des Rehwildes (Capreolus capreolus L.). Folia Zool 83:209–222
Langvatn R (1977) Criteria of physical condition, growth and
development in Cervidae,—suitable for routine studies. Nordic
Council for Wildlife Research, Stockholm
Langvatn R, Loison A (1999) Consequences of harvest on age
structure, sex ratio and population dynamics of red deer in
central Norway. Wildl Biol 5:213–223
Langvatn R, Albon SD, Burkey T, Clutton-Brock TH (1996) Climate, plant phenology and variation in age at first reproduction
in a temperate herbivore. J Anim Ecol 65:653–670
Langvatn R, Mysterud A, Stenseth NC, Yoccoz NG (2004) Timing
and synchrony of ovulation in red deer constrained by short
northern summers. Am Nat 163:763–772
Loe LE, Mysterud A, Langvatn R, Stenseth NC (2003) Decelerating and sex-dependent tooth wear in Norwegian red deer.
Oecologia 135:346–353
Loe LE, Meisingset E, Mysterud A, Langvatn R, Stenseth NC
(2004) Phenotypic and environmental correlates of tooth
eruption in red deer (Cervus elaphus). J Zool 262:83–89
McCullagh P, Nelder JA (1989) Generalized linear models. Chapman and Hall, London
Mitchell B (1967) Growth layers in dental cement for determining
the age of red deer (Cervus elaphus L.). J Anim Ecol 36:279–293
Mysterud A, Yoccoz NG, Stenseth NC, Langvatn R (2000) Relationships between sex ratio, climate and density in red deer: the
importance of spatial scale. J Anim Ecol 69:959–974
Mysterud A, Langvatn R, Yoccoz NG, Stenseth NC (2001a) Plant
phenology, migration and geographic variation in body weight
of a large herbivore: the effect of a variable topography. J Anim
Ecol 70:915–923
Mysterud A, Stenseth NC, Yoccoz NG, Langvatn R, Steinheim G
(2001b) Nonlinear effects of large-scale climatic variability on
wild and domestic herbivores. Nature 410:1096–1099
Mysterud A, Yoccoz NG, Stenseth NC, Langvatn R (2001c) The
effects of age, sex and density on body weight of Norwegian red
deer: evidence of density-dependent senescence. Proc R Soc
Lond Ser B 268:911–919
Pérez-Barberı́a FJ, Gordon IJ (1998) Factors affecting food communution during chewing in ruminants: a review. Biol J Linnean Soc 63:233–256
R Development Core Team (2004) R: a language and environment
for statistical computing. R Foundation for Statistical Computing, Vienna, Austria
Reimers E, Nordby Ø (1968) Relationships between age and tooth
cementum layers in Norwegian reindeer. J Wildl Manage
32:957–961
Skogland T (1988) Tooth wear by food limitation and its life history consequences in wild reindeer. Oikos 51:238–242
Venables WN, Ripley BD (1999) Modern Applied Statistics with Splus. Springer, Berlin Heidelberg New York
Yoccoz NG, Mysterud A, Langvatn R, Stenseth NC (2002) Density-dependent and age related reproductive effort in male red
deer. Proc R Soc Lond Ser B 269:1523–1528
Zima J (1988) Incidence of dental anomalies in Capreolus capreolus
from Czechoslovakia. Folia Zool 37:129–144